Add weakref Python types

[SuperJappie08]

#3134 Add PyWeakRef, PyWeakProxy and PyWeakCallableProxy types, corresponding to Python's weakref.ReferenceType, weakref.ProxyType and weakref.CallableProxyType.

[SuperJappie08]

There are some issues with the new run_bound.
I will resolve these issues now.

[codspeed-hq]

CodSpeed Performance Report

Merging #3835 will not alter performance

_{Comparing SuperJappie08:weakref (7e9aad8) with main (388d176)}

Summary

✅ 67 untouched benchmarks

[davidhewitt]

Thank you and sorry for the slow review, I will try my best to finally look at this tomorrow...

[SuperJappie08]

Thank you and sorry for the slow review, I will try my best to finally look at this tomorrow...

No problem, completely understandable.

I also ran into one last issue, PyPy does not include all the required bindings for the weakref module in their C-api.
At least the bindings required in the way it is done now.

The type-objects for the reference types are not exported (not bound to in pyo3-ffi and I can't find them in the PyPy header files). Also, the memory layout is missing, since in PyPy weakref is not actually a binary module.
That last one should at least be solvable, but I cannot think of a solution for the missing type objects

[davidhewitt]

Thank you very much for this!

It looked like a big diff but that's because you've included all docs and tests, thank you 🙏

I've given this a full read; I have a few initial things which need discussing and changing. I think once those are decided upon & done I might need to take another final read through just due to the size of the diff.

[davidhewitt]

Regarding PyPy, for now I suggest just not offering these types on PyPy.

[SuperJappie08]

I already see a documentation build failure in the CI.

I use a full path to PyNone (crate::types::none::PyNone), which it doesn't like.
Could this be because crate::types::none is private?

(Than the fix would be a small commit, changing crate::types::none::PyNone to crate::types::PyNone)

[SuperJappie08]

It should be fixed now.

I have looked into the Netlify error, it happens on pyo3-ffi which is not changed by this PR, so that error will indeed be resolved by merging.

[SuperJappie08]

Some tests are failing, I will change them to work with the new changes

[SuperJappie08]

It should now work (I have now also tested on Python 3.13)

[davidhewitt]

Let's have another go 👍

[davidhewitt]

I'm going to retry again - it seems to me like it's just cancelling jobs for no clear reason, which is odd.

[davidhewitt]

Ok, this time it failed with a hypothesis flake which I've not seen before. I've never experienced the CI so unreliable 😢

[SuperJappie08]

It's always graalpy-macos that fails (I think).

I find it suspicious...

[davidhewitt]

Let's retry; other stuff is merging ok since. If this still struggles to merge then I think more investigation here of what on this branch is causing the issue.